Amazon has introduced Nova Act, an open-source software development kit designed to create intelligent agents capable of navigating the web much like a human would. While traditional web scraping relies on rigid code that breaks whenever a website changes its layout, Nova Act uses natural language instructions to guide its browser interactions. This flexibility allows agents to adapt to dynamic content, such as rotating banners or promotional pop-ups, ensuring that data extraction remains reliable even as websites evolve.

The system is particularly effective for competitive price intelligence, where retail teams must track thousands of products across competitor sites. Instead of manual spreadsheets, developers can now deploy parallel browser sessions that search for products, extract pricing via structured data models, and handle complex navigation steps simultaneously. This agentic approach significantly reduces the time-to-insight, allowing businesses to adjust their pricing strategies in near real-time rather than relying on stale reports.

Nova Act also integrates human-in-the-loop features to maintain ethical and functional standards. When the agent encounters a Captcha—a security test used to distinguish humans from bots—it can pause and alert a human operator to solve it manually. This hybrid workflow, combined with specialized extensions for popular code editors, simplifies the process of building and scaling automated web agents within the existing AWS ecosystem.

Amazon has introduced Nova Act, an open-source software development kit designed to create intelligent agents capable of navigating the web much like a human would. While traditional web scraping relies on rigid code that breaks whenever a website changes its layout, Nova Act uses natural language instructions to guide its browser interactions. This flexibility allows agents to adapt to dynamic content, such as rotating banners or promotional pop-ups, ensuring that data extraction remains reliable even as websites evolve.

The system is particularly effective for competitive price intelligence, where retail teams must track thousands of products across competitor sites. Instead of manual spreadsheets, developers can now deploy parallel browser sessions that search for products, extract pricing via structured data models, and handle complex navigation steps simultaneously. This agentic approach significantly reduces the time-to-insight, allowing businesses to adjust their pricing strategies in near real-time rather than relying on stale reports.

Nova Act also integrates human-in-the-loop features to maintain ethical and functional standards. When the agent encounters a Captcha—a security test used to distinguish humans from bots—it can pause and alert a human operator to solve it manually. This hybrid workflow, combined with specialized extensions for popular code editors, simplifies the process of building and scaling automated web agents within the existing AWS ecosystem.